The sulfur atoms of the substrate CoA and the catalytic cysteine are required for a productive mode of substrate binding in bacterial biosynthetic thiolase, a thioester-dependent enzyme Gitte Merila ¨ inen 1,2 , Werner Schmitz 3 , Rik K. Wierenga 1,2 and Petri Kursula 1 1 Department of Biochemistry, University of Oulu, Finland 2 Biocenter Oulu, University of Oulu, Finland 3 Biozentrum der Universita ¨ t, Wu ¨ rzburg, Germany The reaction mechanisms of many enzymes depend on thioester chemistry. For example, enzymes involved in lipid metabolism, functioning in both degradative and synthetic pathways, use as substrates fatty acid molecules, conjugated via a reactive thioester moiety to the SH group of pantetheine. This pantetheine moiety is part of either CoA or acyl carrier protein (ACP). All members of the thiolase superfamily of enzymes, including, for example, thiolases as well as the related 3-ketoacyl-ACP-synthases (KAS) [1], have a reactive cysteine in the active site, playing a key role in the reaction cycle by accepting the fatty-acyl moiety from either acyl-CoA (thiolases) or from acyl-ACP (KAS). The kinetic and structural properties of the bacterial Zoogloea ramigera thiolase have been studied in detail [2–11]. This biosynthetic thiolase is a condensing enzyme that catalyses the formation of acetoace- tyl(AcAc)-CoA from two molecules of acetyl(Ac)-CoA, utilizing the unique chemistry of thioester compounds. This reaction consists of two chemical conversions via a ping-pong mechanism [12]: an acetyl transfer and a Keywords active site; calorimetry; coenzyme A; thiolase; X-ray crystallography Correspondence P. Kursula, Department of Biochemistry, University of Oulu, PO Box 3000, FIN-90014, Oulu, Finland Fax: +358 8 5531141 E-mail: petri.kursula@oulu.fi (Received 1 July 2008, revised 2 September 2008, accepted 10 October 2008) doi:10.1111/j.1742-4658.2008.06737.x Thioesters are more reactive than oxoesters, and thioester chemistry is important for the reaction mechanisms of many enzymes, including the members of the thiolase superfamily, which play roles in both degradative and biosynthetic pathways. In the reaction mechanism of the biosynthetic thiolase, the thioester moieties of acetyl-CoA and the acetylated catalytic cysteine react with each other, forming the product acetoacetyl-CoA. Although a number of studies have been carried out to elucidate the thio- lase reaction mechanism at the atomic level, relatively little is known about the factors determining the affinity of thiolases towards their substrates. We have carried out crystallographic studies on the biosynthetic thiolase from Zoogloea ramigera complexed with CoA and three of its synthetic analogues to compare the binding modes of these related compounds. The results show that both the CoA terminal SH group and the side chain SH group of the catalytic Cys89 are crucial for the correct positioning of substrate in the thiolase catalytic pocket. Furthermore, calorimetric assays indicate that the mutation of Cys89 into an alanine significantly decreases the affinity of thiolase towards CoA. Thus, although the sulfur atom of the thioester moiety is important for the reaction mechanism of thioester- dependent enzymes, its specific properties can also affect the affinity and competent mode of binding of the thioester substrates to these enzymes. Abbreviations Ac, acetyl; AcAc, acetoacetyl; ACP, acyl carrier protein; CT, cytosolic thiolase; ITC, isothermal titration calorimetry; KAS, 3-ketoacyl-ACP- synthase; MPD, 2-methyl-2,4-pentanediol; PDB, Protein Data Bank; PP, pantetheine-11-pivalate. 6136 FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS Claisen condensation (Fig. 1). During acetyl transfer, the C1 atom of Ac-CoA reacts electrophilically with the reactive SH of Cys89, forming a covalent acety- lated intermediate [13]. In the subsequent Claisen con- densation, the C2 atom of the second Ac-CoA attacks the Ac-enzyme intermediate nucleophilically; the nucle- ophile is generated by proton abstraction from the C2 of Ac-CoA (Fig. 1). It has been shown that two oxyanion holes in the catalytic cavity play a key role in this reaction mechanism [2]. Oxyanion hole I is formed by a water molecule, referred to as Wat82, and Ne2(His348). This oxyanion hole binds the Ac-CoA thioester oxygen atom, facilitating the nucleophilic attack of the C2 atom of Ac-CoA to the carbonyl carbon atom of the acetyl moiety of the acetylated enzyme. The electrophilic reactivity of the latter atom Fig. 1. The thiolase reaction mechanism and the compounds used. (A) The thiolase reaction. In the biosynthetic direction, the overall reac- tion uses two molecules of Ac-CoA to generate AcAc-CoA and CoA. Cys89 is activated for nucleophilic attack by His348. (B) Comparison of the covalent structures of CoA (top), SPP (middle) and OPP (bottom). SPP is different from CoA, such that the pantetheine moiety has an ester linkage to a pivalate group instead of 3¢-phospho-ADP. In OPP, the reactive SH group of SPP is further replaced by an OH group. In CoA, certain atoms of the pantetheine moiety are numbered. G. Merila ¨ inen et al. Sulfur interactions at the thiolase active site FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS 6137 is increased by oxyanion hole II, being formed by N(Cys89) and N(Gly380) and binding the thioester oxygen atom of the Cys89-bound acetyl group; oxyan- ion hole II is similar to the ‘classical’ oxyanion hole seen in, for example, serine proteases. Although mutation of the reactive Cys89 to serine still allows for transacetylation at low efficiency [6], the exchange of the substrate thioester sulfur atom to an oxygen apparently prevents the reaction from taking place [4–7]. The compounds used in such studies have consisted of thio- and oxoesters of a CoA analogue, pantetheine-11-pivalate (PP) [3], which lacks the 3¢-phospho-ADP moiety of CoA, having a pivalate group at the 11-hydroxyl moiety of pantetheine instead (Fig. 1). The acetyl and acetoacetyl thioesters of PP (Ac-SPP and AcAc-SPP) have been found to be substrates of the bacterial biosynthetic thiolase [3–5]. Furthermore, 3-pentynoyl-SPP was used to identify Cys378 as a catalytic residue [7,8]. The oxoesters of PP are nonreactive [5–7,14]. Lower reactivity of the oxoester compared to the thioester is also seen when comparing the kinetic properties of crotonyl-CoA and crotonyl-oxyCoA: the hydration rate of the oxoester by crotonase is 330-fold lower [15]. Due to different resonance properties of the thio- and oxoesters [16], the pK a of the a carbon of a thio- ester is approximately 21, whereas the pK a for the corresponding atom in an oxoester is 26 [17]. This will make Claisen condensation more difficult with an oxo- ester because the nucleophilic carbanion is harder to generate (Fig. 1). Furthermore, the reactivity of the carbonyl carbon of thioesters and oxoesters towards nucleophilic attack is different [18,19], with oxoesters being less reactive. Thus, from the chemical properties of oxoesters, it can be postulated that they are less reactive in both halves of the thiolase reaction than thioesters. No previous studies have addressed the pos- sibility that the poor reactivity of the esters of OPP in the thiolase reaction could also be, at least partially, related to the preferred nonproductive binding modes of these oxoesters. Members of the thiolase superfamily, being struc- turally and mechanistically related [1], are of consider- able interest in the field of biotechnology; the applications utilizing their potential in biosynthetic reactions are diverse [20–25]. The Z. ramigera thiolase is a very close homologue of the human cytosolic thiolase (CT) [11], which is a key enzyme in the cho- lesterol synthesis pathway [26–28], and a detailed understanding of the binding determinants of its substrate could also help in the development of suit- able drugs towards CT, with the aim of lowering high cholesterol levels. Additionally, other human thiolases have been found to be drug targets for the treatment of heart failure [29–32]. In the present study, we used a combination of crys- tallography and calorimetry to analyse the binding determinants of CoA to the bacterial biosynthetic thio- lase from Z. ramigera, concentrating specifically on the SH groups of the substrate and the catalytic Cys89, which is conserved throughout the thiolase superfam- ily. The results obtained indicate that the sulfur atoms of both the enzyme and the substrate are important for the correct productive mode of binding of CoA in the thiolase active site, improving our understanding of substrate recognition by thioester-dependent enzymes. Results and Discussion The biosynthetic thiolase from Z. ramigera is a tetra- meric 160 kDa enzyme, consisting of four identical subunits of 392 residues (Fig. 2). Three domains of approximately equal lengths have been identified in the thiolase fold. The core of the monomer is formed by the N-terminal domain (with the catalytic cysteine) and the C-terminal domain. The third domain, referred to as the loop domain, protrudes out of the N-terminal domain. The loop domain covers the catalytic site and provides the binding site for the 3¢-phospho-ADP moiety of CoA (Fig. 2). In the present study, we aimed to analyse in detail, using X-ray crystallography and isothermal titration calorimetry, the factors that influence the productive mode of binding of a CoA substrate to the active site of biosynthetic thiolase. We solved five new liganded crystal structures of Z. ramigera thiolase (Table 1; see also Fig. S1) and compared these with the previously available structures of liganded complexes of this and other thiolases. The results obtained, as discussed in detail below, indicate a crucial role for sulfur–sulfur interactions in defining the catalytically competent binding mode of CoA in the active site. The pantetheine binding tunnel of biosynthetic thiolase When CoA binds to Z. ramigera biosynthetic thiolase, its pantetheine moiety enters the narrow pantetheine binding tunnel and its SH group closes off the catalytic cavity, near the catalytically important residues Cys89, His348 and Cys378. The residues forming the walls of the tunnel include the side chains of Leu148, His156, Met157, Ala234, Phe235, Ala243, Ala246, Ser247, Gly248 and Leu249 of the loop domain, as well as Ala318 and Phe319 from the C-terminal domain Sulfur interactions at the thiolase active site G. Merila ¨ inen et al. 6138 FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS b-strand Cb2, and Met134 of the neighboring subunit. These residues encircle the pantetheine moiety. The two peptide moieties of the pantetheine unit are both similarly tightly wedged between Phe235 and Leu249 (the outermost peptide bond, near the bulk solvent) and between Phe319 and Leu148 (the innermost pep- tide bond, near the catalytic site). The atoms shaping the binding environment of the terminal sulfur atom of CoA are listed in Table 2 and shown in Fig. 2. Apart from the sulfur-containing resi- dues Cys89, Met157, Met288 and Cys378, these also include side chain atoms of Ala318 and Phe319; the latter two residues are a part of the highly conserved NEAF sequence motif [1]. Within 4.8 A ˚ from the CoA sulfur atom, there are also the catalytic water (Wat82) and Ne2(His348) (Table 2) (i.e. the atoms forming oxyanion hole I). Wat82 is hydrogen bonded to Nd1(Asn316) and Wat49. Wat82 and Wat49 are pres- ent in each of the five new structures (Fig. 3). A prominent feature of the binding pocket for the CoA sulfur atom is the presence of the four sulfur atoms from Cys89, Met157, Met288 and Cys378. Due to the high polarizability of sulfur atoms [33], it is expected that sulfur–sulfur interactions will contribute significantly to the van der Waals binding energy. Cys89 and Cys378 are catalytic residues, but also Met157 and Met288 are highly conserved in the thio- lase family. An interesting exception is thiolase T2; in this case, Met288 is replaced by a phenylalanine. This is a unique feature of the T2 thiolase sequence, which correlates with its unique substrate specificity: T2 is able to use not only acetoacetyl-CoA, but also the branched 2-methylacetoacetyl-CoA molecule as a substrate [34]. AB C Fig. 2. The active site of thiolase. (A) The overall shape of the biosynthetic thiolase tetramer. The four individual active sites of the tetramer are indicated by the bound CoA molecules. The catalytic site of the yellow domain is marked by an arrowhead. (B) The binding mode of CoA to the thiolase monomer. The three domains are coloured yellow (N-terminal domain), light-green (loop-domain) and light gray (C-terminal domain), respectively. The NEAF motif (including Phe319) is coloured red, the loop 231–240 (shifted in the OPP ⁄ Ac-OPP structures and con- taining Phe235) is coloured purple and the catalytic Cys89 is coloured dark blue. His156 at the entrance of the pantetheine-binding cavity is coloured orange. Note how the CoA interacts mainly with the loop domain and Cys89. (C) A detailed view of the surroundings of the termi- nal moiety of CoA in the thiolase active site, as seen in the complex of unmodified thiolase with CoA (a similar view to that in B). Contacts of the terminal sulfur are coloured green, hydrogen bonds are coloured red, and interactions between hydrophobic side chains and the planar amide bonds of CoA are coloured yellow. G. Merila ¨ inen et al. Sulfur interactions at the thiolase active site FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS 6139 Only two hydrophilic side chains point into the pantetheine binding tunnel: Ser247 and His156. Ser247 is known to adopt two different conformations; in the unliganded form, it points away from the tunnel, and, in the liganded conformation, it points into the tunnel, interacting with the N4 atom of the CoA substrate [2,10]. His156 has previously been observed in only one conformation, having van der Waals contacts to the pantetheine moiety of CoA at the outer edge of the tunnel. Water molecules are also present in the pantetheine binding tunnel, both in the unliganded and liganded states [2,11]. There is one conserved water near the bottom of the pantetheine binding tunnel (Wat149), interacting with O(Gly248) and O(His348). Table 2. Distances between the terminal sulfur or oxygen atom of the pantetheine moiety of the active site ligand and surrounding atoms. As a reference, the structure of the unmodified Z. ramigera thiolase complexed with CoA (PDB entry 1DLV) has been used, considering all atoms within 4.8 A ˚ of the CoA sulfur. The thiolases are from Z. ramigera, unless otherwise specified. In the oxidized structures, the active site cysteine, Cys89 or its equivalent, has been oxidized to a sulfenic acid. In the acetylated structures, the active site cysteine has been acetylated. Thiolase Wild-type C89A Wild-type Wild-type oxidized Wild-type oxidized Human CT oxidized Human T2 oxidized Human T2 acetylated Wild-type acetylated Ligand CoA CoA SPP OPP CoA CoA CoA CoA CoA Sc-Cys89 3.9 – 4.1 5.2 3.8 4.2 4.3 3.8 4.3 Sc-Cys378 4.7 4.3 4.9 6.2 5.2 4.7 4.9 5.1 4.7 Sd-Met157 4.1 5.7 4.3 7.6 3.9 4.1 4.1 4.3 4.1 Cc-Met157 3.7 5.5 3.8 6.5 3.2 3.8 3.8 3.9 3.7 Cb-Met157 4.0 5.8 3.8 6.2 3.6 3.8 3.8 4.2 3.7 Sd-Met288 4.4 4.7 4.6 7.4 4.4 4.6 – – 4.5 Ce-Met288 3.7 4.3 3.4 6.2 3.7 3.6 – – 3.4 Ce2-Phe319 4.2 5.3 3.6 4.4 4.1 3.9 3.6 4.1 3.9 Cb-Ala318 4.2 3.8 3.9 3.2 4.5 3.8 3.8 4.0 4.0 Ne2-His348 4.6 3.3 4.9 3.5 4.9 4.9 5.2 5.0 4.7 Wat82 4.8 3.5 5.1 5.1 5.3 4.9 4.8 4.8 4.9 PDB entry 1DLV 2VTZ 2VU2 2VU1 2VU0 1WL4 2IBU 2F2S 1QFL Table 1. Data processing and refinement statistics. The numbers in parentheses refer to the highest resolution shell. Complex SPP OPP (oxidized Cys89) CoA (oxidized Cys89) CoA (C89A) Ac-OPP (oxidized Cys89) Beamline EMBL ⁄ DESY X13 EMBL ⁄ DESY BW7B EMBL ⁄ DESY X11 EMBL ⁄ DESY X13 Rotating anode Wavelength (A ˚ ) 0.81 0.84 0.81 0.81 1.5418 Data processing Resolution range (A ˚ ) 20–2.65 (2.72–2.65) 20–1.51 (1.55–1.51) 20–1.87 (1.95–1.87) 20–2.30 (2.40–2.30) 20–2.07 (2.20–2.07) <I ⁄ rI> 11.8 (3.5) 8.7 (2.1) 12.1 (4.4) 15.6 (4.8) 9.5 (4.5) Completeness (%) 95.2 (98.4) 93.8 (70.1) 97.3 (86.3) 97.2 (86.1) 93.9 (83.6) R merge (%) 9.9 (32.0) 9.1 (54.6) 7.1 (19.0) 6.8 (26.6) 10.0 (23.2) Redundancy 3.2 (2.5) 3.5 (2.8) 3.1 (2.4) 3.5 (2.7) 2.8 (2.2) Wilson B (A ˚ 2 ) 3621 253025 Space group P2 1 P2 1 P2 1 P2 1 P2 1 Unit cell parameters (A ˚ , °) 84.3, 79.2, 150.8 90, 92.9, 90 84.3, 78.7, 148.3 90, 92.9, 90 84.4, 79.1, 148.8 90, 92.7, 90 84.2, 79.6, 148.9 90, 92.1, 90 84.4, 79.0, 148.8 90, 93.0, 90 Refinement R cryst (%) 23.1 21.6 19.4 22.2 16.1 R free (%) 28.6 24.9 22.9 26.2 21.2 rmsd Bond distances (A ˚ ) 0.011 0.014 0.018 0.009 0.014 rmsd Bond angles (°) 1.2 1.4 1.6 1.1 1.4 rmsd B factors for bonded atoms (main chain, side chain) (A ˚ 2 ) 1.0,2.0 1.5,1.7 1.4,2.7 0.7,1.3 1.9,2.9 PDB entry 2vu2 2vu1 2vu0 2vtz 1ou6 Sulfur interactions at the thiolase active site G. Merila ¨ inen et al. 6140 FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS This water is present in all of the new structures (Fig. 3). The mode of binding of SPP closely resembles that of CoA The crystal structure of Z. ramigera thiolase was solved in complex with SPP, a functional CoA ana- logue. SPP and its thioesters are known to be func- tional substrates for thiolase, but kinetic constants have been reported only for AcAc-SPP [3]; for exam- ple, K m and k cat are 73 lm and 469 ⁄ s for AcAc-SPP, whereas they are 24 lm and 465 ⁄ s, respectively, for AcAc-CoA. In the crystal structure, the binding modes of the reactive sulfur moieties of SPP and CoA are highly similar, explaining the reactive nature of SPP. For example, the distance between the sulfur atoms of SPP and the reactive Cys89 is 4.1 A ˚ , whereas the cor- responding distance for the CoA complex is 3.9 A ˚ (Table 2). The pantetheine moiety of SPP binds in the pant- etheine binding tunnel in a similar (but not identical) way to that seen for CoA (Fig. 3A). The pivalate head group points outwards from the pantetheine binding tunnel, overlapping with the pyrophosphate moiety of CoA. It lies on a hydrophobic surface comprising Leu249, Phe18 and Met134; the latter comes from the tetramerization loop of an opposing subunit. These results indicate that the 3¢-phospho-ADP group of CoA is not crucial for the correct positioning of the reactive end group of the substrate in the thiolase cata- lytic cavity. However, it could be important for increasing the affinity of binding, as well as the solub- ility of the substrate. Interestingly, the side chain of Ser247 points away from SPP, as previously seen in unliganded thiolase; in the case of a bound CoA ligand, it points towards the ligand. This difference correlates with minor conforma- tional differences between the pantetheine moieties of CoA and SPP. The highly similar mode of binding of the reactive moieties of CoA and SPP in the active site of thiolase provides a structural basis for the observa- tion that SPP compounds are functional substrates of thiolase. The binding mode of OPP is unproductive The structure of Z. ramigera biosynthetic thiolase was also solved in the presence of OPP and its acetylated analogue. The complex with OPP was refined at a res- olution of 1.51 A ˚ . The structure indicates a surprising binding mode (Fig. 3A); OPP is bound further away from the catalytic cavity than SPP and CoA. A B C Fig. 3. Comparison of the active site ligand binding modes to bio- synthetic thiolase. (A) The binding modes of SPP (magenta), OPP (cyan), Ac-OPP (orange) and CoA (gray). A water molecule (I) in the OPP complex is bound to oxyanion hole I. His156 has a double con- formation in the Ac-OPP complex. (B) The binding mode of CoA to thiolase harboring a modified active-site Cys89 is identical to that seen for unmodified thiolase. The CoA complexes of unmodified (gray), oxidized (yellow) and acetylated (green) Z. ramigera thiolase; the human oxidized CT CoA complex (dark gray); and the human acetyl-T2 CoA complex (pink) are shown. Oc(Ser247) points towards the pantetheine moiety, except for the complex between oxidized Z. ramigera thiolase and CoA, in which Ser247 has a dou- ble conformation. (C) Superposition of CoA complexes of wild-type (gray) and C89A (brown) thiolase. In the CoA C89A complex, there are two extra water molecules in the active site cavity; one water (II) is bound to oxyanion hole II and another nearby water molecule (e) is hydrogen bonded to it. G. Merila ¨ inen et al. Sulfur interactions at the thiolase active site FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS 6141 The catalytic cavity lies at the bottom of the pant- etheine-binding tunnel. The terminal oxygen of OPP is 5.2, 7.4, 6.2 and 7.6 A ˚ away from the sulfur atoms of Cys89, Met288, and Cys378 and Met157, respectively (Table 2). OPP has a terminal hydroxyl group, which can form hydrogen bonds with water when in solution, and with the protein once bound in the complex. From the structure, it is apparent that the hydroxyl group of OPP is in hydrogen bonding interaction with Ne2(His348) (at 3.2 A ˚ ) and O(Ala318) (at 3.3 A ˚ ), being inserted into a pocket between these atoms. This pocket would be too tight for a sulfur atom, such as those in CoA and SPP. This mode of binding places the terminal oxygen atom further away from the center of the catalytic cavity and, in this structure, oxyanion hole I has a water molecule bound (Fig. 3). In the CoA mode of binding [10], there is a water- mediated hydrogen bond between Ne2(His156) and O9(CoA), as well as a hydrogen bond between Oc(Ser247) and N4(CoA), and the carbonyl O(Ser247) is hydrogen bonded to N8(CoA). For OPP, these hydrogen bonds are not observed. The side chain of Ser247 points away from the bound OPP, and there only exists a hydrogen bond between O(Ser247) and N4(OPP) because the pantetheine group has moved away from the bottom of the tunnel compared to SPP and CoA (Fig. 3). Consequently, the terminal t-butyryl moiety of OPP is in a different conformation than that of SPP. Because this group is in a different conforma- tion in SPP and OPP, it is not specifically recognized by thiolase, which is as expected for a synthetic ana- logue. The t-butyryl group of OPP is bound in a pocket created by a rotation of the His156 side chain around v1, away from its position observed in all earlier structures. The residues forming the walls of this hydrophobic pocket include Met143, Ile144, Leu148, His156, Ala234, Phe235, Leu249 and Met134 (from the opposing subunit). The side chains of Ile144 and Met134 are also in a slightly different conforma- tion compared to all previous structures. Moreover, the entire loop containing residues 231–240 (Fig. 2) moves away by approximately 1 A ˚ in the presence of OPP; at the same time, the B factors for this loop indi- cate a more rigid structure than in the presence of CoA (data not shown). These changes are relatively extensive compared to those seen upon CoA binding, in which case it has been noted that the Sc atom of Cys89 and the Oc atom of Ser247 are the only non- hydrogen protein atoms that move detectably during the thiolase reaction cycle [2,9,10]. OPP differs from SPP by only one atom, and the replacement of the sulfur atom of SPP by an oxygen changes the interaction such that OPP binds in a differ- ent way to the thiolase active site. This mode of binding of OPP to biosynthetic thiolase is unproductive, and not competent for catalysis. Nonproductive binding is a well-known phenomenon in enzymological studies [33]. Classic examples from structural studies include the binding of (NAG) 3 to lysozyme [35] and that of the ind- olyl acryloyl moiety in the active site of chymotrypsin [36]. More recent examples concern crystallographic binding studies of elastase [37], dihydrofolate reductase [38] and methylmalonyl CoA decarboxylase [39]. We also solved the structure of the complex of thio- lase with Ac-OPP; from the electron density map and the subsequent structure comparison, it can be seen that the binding mode is the same as that for OPP; furthermore, the structural changes of the protein part described for the OPP complex are seen as well in the Ac-OPP complex. Due to low occupancy of the com- pound in the structure, further detailed analysis of the exact binding mode has not been performed, but the mode of binding of the acetyl moiety is clearly differ- ent from the mode of binding of the acetyl moiety of Ac-CoA. Apparently, the common carbonyl oxygen atom of the acetyl group of these two molecules does not provide enough binding energy to favour a similar mode of binding for Ac-OPP and Ac-CoA. Binding mode of CoA to thiolases oxidized or acetylated at Cys89 The structure of CoA-complexed Z. ramigera thiolase, in which the catalytic Cys89 was oxidized, was also determined. This was carried out to allow a detailed comparison of the binding mode of CoA to thiolase when the sulfur atom of Cys89 is modified covalently. The oxidation of the catalytic cysteine to a cysteine sulfenic acid has been observed in complexes of the human cytosolic thiolase [11] and the human mito- chondrial thiolase T2 [34]. The extra oxygen atom points away from the ligand binding pocket into oxy- anion hole II and, in these complexes, the mode of binding of CoA is unaffected by the different oxidation state of the cysteine sulfur (Fig. 3B). The same applies to thiolases complexed with CoA in their acetylated form, as seen in the structures of human T2 [Protein Data Bank (PDB) entry 2F2S] and Z. ramigera thio- lase [9]. For the Z. ramigera thiolase, the mode of binding of CoA is unaffected by the oxidation (or acet- ylation) state of the active site cysteine (Fig. 3B). Indeed, the mode of binding of the CoA sulfur atom is conserved also in the structures with acetyl-CoA com- plexed with the acetylated active site [10]. In the OPP complexes, Cys89 is oxidized to a cysteine sulfenic acid, but the above results indicate that the different Sulfur interactions at the thiolase active site G. Merila ¨ inen et al. 6142 FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS modes of binding of OPP and Ac-OPP, when com- pared with the corresponding SPP or CoA compounds, cannot be attributed to the oxidation state of the cata- lytic cysteine. Binding mode of CoA to the C89A thiolase variant To better understand the importance of the sulfur– sulfur interactions in the catalytic cavity of thiolase, the CoA binding properties of its C89A variant were also studied. In the crystal structure, CoA binding essentially follows the wild-type mode, apart from the reactive terminus of CoA. The terminal sulfur is inserted deeper into the catalytic cavity by approxi- mately 1.7 A ˚ , being sandwiched between His348 and Cys378. In the wild-type complex, the reactive termi- nus of CoA is apparently in a somewhat strained con- formation (Fig. 3C). This indicates that the sulfur atom of Cys89 is crucial for the correct positioning of the reactive moiety of CoA. In addition, in the com- plex of the C89A thiolase mutant with CoA, a water molecule is placed in oxyanion hole II, formed by the backbone nitrogens of residues Cys89 and Gly380, and a second water is observed hydrogen bonded to O(Gly147) (Fig. 3C). The tight fit of CoA into the active site in this case is illustrated by the distances from the S atom of CoA to Ne2(His348) (3.3 A ˚ ), Cb(Ala89) (3.6 A ˚ ), a water bound to oxyanion hole II (3.3 A ˚ ) and water 82 (3.5 A ˚ ). The distances to the active site sulfurs are: Sc(Cys378) (4.3 A ˚ ), Sd(Met288) (4.7 A ˚ ) and Sd(Met157) (5.7 A ˚ ), reflecting the overall weakening of the sulfur–sulfur van der Waals interac- tions in the CoA binding mode of the C89A variant, especially concerning Met157 (Table 2), in addition to the absence of the sulfur–sulfur interaction between CoA and Cys89. Calorimetric analysis of the binding of CoA by wild-type thiolase and the C89A variant The detailed crystallographic binding studies described above show that the sulfur–sulfur interactions in the active site of thiolase are important for competent binding of the substrate. Therefore, a calorimetric study was carried out to further investigate the impor- tance of sulfur–sulfur interactions for the affinity of thiolase towards its substrate. Isothermal titration calorimetry (ITC) was used to compare the affinity of CoA to wild-type thiolase and its C89A variant. The results obtained (Fig. 4 and Table 3) indicate that the affinity is significantly higher for the wild-type enzyme compared to the C89A variant, although only one sulfur atom is missing in the mutant. The differ- ence in affinity corresponds to a loss of free energy of binding (DDG) of 0.8 kcalÆmol )1 , whereas the loss of enthalpy of binding (DDH) is 3.7 kcalÆmol )1 . The key difference in the crystal structures is the absence of the sulfur atom of Cys89, generating small structural dif- ferences (Fig. 3); the binding cavity for the reactive moiety of the substrate is less compact in the C89A mutant, which is consistent with the favorable differ- ence binding entropy term [D(-TDS)]. The magnitude of the unfavorable DDH term indicates that the sulfur– AB Fig. 4. Calorimetric analysis of CoA binding by (A) wild-type and (B) C89A thiolase. Curve fitting in both cases was carried out by setting the binding stoichiometry to 1. Note the different scales on the y-axis, which are related to the binding enthalpy. G. Merila ¨ inen et al. Sulfur interactions at the thiolase active site FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS 6143 sulfur interactions provide a significant contribution to the binding energy of CoA [33]. Concluding remarks Cys89 is the catalytic residue in Z. ramigera thiolase, and its mutation to alanine leads to a complete loss of activity [2]. A corresponding catalytic cysteine residue is conserved in all members of the thiolase superfamily. In the present study, we have shown that, apart from the lack of activity, the mutation C89A also lowers the affinity of biosynthetic thiolase towards CoA signifi- cantly. In complexes between biosynthetic thiolase and CoA or SPP, a functional substrate analogue, the sulfur atom of the substrate is always closely embraced by four sulfur atoms from the enzyme. When OPP is used, being identical to SPP except for the replacement of the terminal sulfur atom by oxygen, the binding mode of the ligand also changes, resulting in a nonproductive binding mode. Our data indicate an important role for the interactions between the CoA substrate sulfur group and the thiolase active site in assuring an optimal affinity, as well as a competent mode of binding. There exists considerable interest in the properties and engineering possibilities of enzymes in the thiolase superfamily, both within biotechnology and pharma- cology, due to their involvement in, for example, dif- ferent natural product synthesis pathways, as well as in lipid and cholesterol metabolism. The results obtained in the present study indicate that the sulfur atom of the thioester moiety is not only important because of the high intrinsic reactivity of thioester sub- strates, but also that it can play a key role in achieving the proper affinity and competent mode of binding of the substrates in the active site cavities of thioester- dependent enzymes. This should be taken into consid- eration when designing, for example, new substrates or inhibitors for these enzymes. Experimental procedures Protein expression, purification and crystallization Z. ramigera thiolase and its C89A mutant were expressed and purified as previously described [2,9]. Crystallization of wild-type thiolase and its C89A variant were carried out at 22 °C using vapour diffusion in a mother liquour contain- ing 1 m Li 2 SO 4 , 0.9 m (NH 4 ) 2 SO 4 , 0.1 m sodium citrate (pH 5), 1 mm EDTA, 1mM NaN 3 and 1 mm dithiothreitol. Synthesis of OPP and SPP All intermediates and products were purified using silica gel chromatography, and their purity was checked by TLC. The identity of the compounds was demonstrated by 1 H-NMR. All chemicals were obtained from Sigma-Aldrich (Taufkirchen, Germany). Synthesis of ethanolamino-tert.butyl-diphenyl-silane A solution of 10 mmol tert.butyl-diphenyl-silyl-chloride in 20 mL of tetrahydrofurane was added dropwise with stirring to 50 mmol ethanolamine in 30 mL of tetrahydrofurane at 0 °C. After stirring for 2 h at room temperature (RT), the solution was evaporated. Yield: 8.7 mmol ethanolamino- tert.butyl-diphenyl-silane. Purification of pantothenoic acid A solution of 20 mmol d(+)-calcium pantothenate in 5 mL of 4 m HCl was extracted once with 40 mL of chloro- form ⁄ methanol (2 : 1) and three times with 40 mL of chlo- roform ⁄ methanol (9 : 1). The organic layers were evaporated. Yield: 10 mmol pantothenoic acid. Synthesis of tert.butyl-diphenyl-silyl-pantetheinate Dicyclohexyl-carbodiimide (11 mmol) was added to 10 mmol pantothenoic acid in 50 mL of tetrahydrofurane. The suspension was stirred until dicyclohexyl-carbodiimide dissolved completely; 8.7 mmol ethanolamino-tert.butyl- diphenyl-silane was then added. The solution was stirred for 4 h at RT. Yield: 2.6 mmol tert.butyl-diphenyl- silyl-pantetheinate. 1 H-NMR: C2: 3.48 p.p.m., s.; C11: 3.75 p.p.m., t.; phenyl: 7.65 p.p.m., d. Synthesis of tert.butyl-silyl-OPP A solution of 4 mmol pyridine was added to a solution of 2 mmol tert.butyl-diphenyl-silyl-pantetheinate in 50 mL of dichlormethane; 2 mmol pivaloylchloride in 20 mL of dic- hlormethane was then added dropwise with stirring. After Table 3. Calorimetric analysis of CoA binding to the Z. ramigera biosynthetic thiolase at 25 °C. The values and error estimates are calculated from separate measurements (three for the wild-type and two for C89A). K a and K d are the association and dissociation constants, respectively. Sample DH (kcalÆmol )1 ) )TDS (kcalÆmol )1 ) DG (kcalÆmol )1 ) K a (M )1 ) K d (lM) Wild-type )6.7 ± 0.9 1.1 ± 0.8 )5.6 ± 0.2 1.3 · 10 4 ± 0.4 · 10 4 81 ± 29 C89A )3.0 ± 0.03 )1.8 ± 0.2 )4.8 ± 0.2 3.2 · 10 3 ± 0.1 · 10 3 307 ± 64 Sulfur interactions at the thiolase active site G. Merila ¨ inen et al. 6144 FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS stirring for 24 h at RT, the residue was dissolved in 20 mL of acetoacetate and extracted with 25 mL each of 0.1 m HCl, saturated CuSO 4 in water, and 2 m NaCl. Yield: 1.3 mmol tert.butyl-silyl-OPP. 1 H-NMR: C2: 3.67 p.p.m., s.; C11: 3.92 p.p.m., t.; phenyl: 7.82 p.p.m., d.; pivalate- methyl: 1.26 p.p.m., s. Synthesis of OPP A solution of 1 mL of hydrogen fluoride was added to 1.3 mmol tert.butyl-silyl-O-pantetheine-11-pivalate in 30 mL of acetonitrile. After stirring for 30 min at RT, the solution was evaporated. The residue was suspended in 20 mL of water and extracted three times with 20 mL of ethylacetate. Yield: 1.2 mmol OPP. 1 H-NMR: C2: 3.78 p.p.m., s.; C11: 4.03 p.p.m., t.; pivalate-methyl: 1.43, s. Synthesis of Ac-OPP A solution of 1 mmol pyridine and 0.5 mmol acetyl chlo- ride was added to a solution of 0.1 mmol OPP in 2 mL of chloroform and stirred at RT for 15 min. The suspension was extracted three times with 2 mL of brine. Yield: 89 lmol Ac-OPP. 1 H-NMR: C2: 3.59 p.p.m., s.; C11: 4.03 p.p.m., t.; pivalate-methyl: 1.23, s.; acetyl-methyl: 2.17 p.p.m., s. Synthesis of SPP Bis(N-pantothenylamidoethyl) disulfide was converted into Bis(N-pantothenylamidoethyl-11-pivalate) disulfide accord- ing to the synthesis of tert.butyl-silyl-OPP. The resulting disulfide was cleaved by treatment with mercaptoethanol. Crystal handling Crystal structures were determined for the complexes of thiolase with OPP, Ac-OPP and SPP. OPP, Ac-OPP and SPP were poorly soluble in aqueous buffer solutions. In the OPP soaking experiment, OPP could be dissolved in the cryoprotectant solution, containing, in addition to the con- stituents of the well solution, 12% 2-methyl-2,4-pentanediol (MPD), 12% glycerol and 100 mm OPP. Prior to data collection, the crystal was transferred to this cryosolution. A second crystal was soaked similarly in a solution contain- ing Ac-OPP instead of OPP. In both soaking experiments, the crystals started suffering during the soak under all tested conditions; thus, the soaking time was approximately 1 min. SPP was dissolved in MPD, at an approximate concen- tration of 100 mm. For soaking SPP into thiolase crystals, the crystals were transferred to drops of mother liquor con- taining one-fifth of the SPP stock solution. This soaking, in approximately 20% MPD and 20 mm SPP in mother liquor, was allowed to continue for 4 days prior to data collection. This experiment was performed for both a wild- type thiolase crystal and a C89A mutant crystal. Analysis of the data collected from these crystals showed the mode of binding of SPP to wild-type thiolase, but no binding was observed when using the C89A crystals. Two more structures were determined and analysed: the structure of a complex of CoA with the C89A variant and a complex of CoA bound in the active site with an oxidized Cys89. The latter structure was obtained from a soaking experiment of a wild-type thiolase crystal with 5 mm b-hy- droxybutyryl-CoA for 1 min. During refinement, it was seen that the structure contained only CoA and the active site Cys89 was oxidized. The soaking of CoA into a C89A thiolase crystal was performed at 5 mm CoA. Data collection, structure solution and refinement Data were collected on beamlines BW7B, X11 and X13 at the EMBL-Hamburg Outstation ⁄ DESY (Hamburg, Germany), except for the data from the Ac-OPP complex, which were collected on a Nonius FR591 rotating anode source (Bruker AXS, Delft, the Netherlands). Data process- ing (Table 1) was carried out with xds [40] and xdsi [41]. Five percent of all reflections were used for calculating the free R factor [42]. The structures were refined using refmac5 [43]. tls parameters [44] were applied, and water molecules were added using arp ⁄ warp [45]. CoA, OPP, Ac-OPP and SPP were built when their electron densities were strong and continuous. Model building and analysis were performed using o [46] and coot [47]. The refinement statistics are given in Table 1. The coordinates and structure factors were deposited to the PDB under the following accession codes: 2VU2 (thio- lase-SPP complex), 2VU1 (thiolase-OPP complex), 2VU0 (oxidized thiolase-CoA complex), 2VTZ (C89A thiolase- CoA complex) and 1OU6 (Ac-OPP complex). Structure analysis The A and B subunits of the biosynthetic thiolase tetra- mer are always best defined in this crystal form of Z. ram- igera thiolase due to the layer-like packing of the thiolase tetramers in the crystal lattice, and the B subunit has been used for previous analyses [2,9,10]. In the case of SPP and OPP, however, the ligand is slightly better defined in the A subunit, and this subunit has mainly been analysed in the present study with respect to these ligands. All dis- cussed features can, however, be seen in the B subunit. Cys89 was oxidized in the structures of the complexes with OPP and Ac-OPP, and it was built as a cysteine sulfenic acid, with the oxygen atom pointing into oxyan- ion hole II. G. Merila ¨ inen et al. Sulfur interactions at the thiolase active site FEBS Journal 275 (2008) 6136–6148 ª 2008 The Authors Journal compilation ª 2008 FEBS 6145 [...]... and degassed The titration was carried out at 25 °C Blank experiments (without protein) were made to estimate the heat of dilution for CoA The binding isotherms obtained by integrating the injection peaks were fitted to appropriate binding models (the stoichiometry was set to 1 per subunit) by using the origin software (MicroCal) The titration curve was fitted by the nonlinear least-squares method, and. .. Acta 1784, 1742–1749 33 Fersht A (1999) Structure and Mechanism in Protein Science: A Guide to Enzyme Catalysis and Protein Folding WH Freeman and Co., New York, NY 34 Haapalainen AM, Merilainen G, Pirila PL, Kondo N, ¨ ¨ Fukao T & Wierenga RK (2007) Crystallographic and kinetic studies of human mitochondrial acetoacetylCoA thiolase: the importance of potassium and chloride ions for its structure and. .. Belgium) and Dr Inari Kursula (University of Oulu, Finland) are gratefully acknowledged This study was supported by the Academy of Finland (grant 200966) 6146 References 1 Haapalainen AM, Merilainen G & Wierenga RK (2006) ¨ The thiolase superfamily: condensing enzymes with diverse reaction specificities Trends Biochem Sci 31, 64–71 2 Kursula P, Ojala J, Lambeir AM & Wierenga RK (2002) The catalytic cycle of. .. (49, 82 and 149) in the text and figures corresponds to earlier analyses of subunit B In two of the structures used for the comparison, the active site cysteine was also oxidized: the CoA complexes of human CT (PDB entry 1WL4) [11] and human mitochondrial thiolase (T2; PDB entry 2IBU) [34] In two other structures of CoA complexes, the active site cysteine was acetylated: the Z ramigera thiolase (PDB... catalytic cycle of biosynthetic thiolase: a conformational journey of an acetyl group through four binding modes and two oxyanion holes Biochemistry 41, 15543–15556 3 Davis JT, Moore RN, Imperiali B, Pratt AJ, Kobayashi K, Masamune S, Sinskey AJ, Walsh CT, Fukui T & Tomita K (1987) Biosynthetic thiolase from Zoogloea ramigera I Preliminary characterization and analysis of proton transfer reaction J Biol... AJ & Masamune S (1991) Biosynthetic thiolase from Zoogloea ramigera Evidence for a mechanism involving Cys-378 as the active site base J Biol Chem 266, 8369–8375 8 Williams SF, Palmer MA, Peoples OP, Walsh CT, Sinskey AJ & Masamune S (1992) Biosynthetic thiolase from Zoogloea ramigera Mutagenesis of the putative active-site base Cys-378 to Ser-378 changes the partitioning of the acetyl S -enzyme intermediate.. .Sulfur interactions at the thiolase active site G Merilainen et al ¨ For the analysis and comparisons of the new structures, six additional structures were used The structures were superimposed on each other using the ssm [48] approach implemented in coot [47] The sequence numbering always refers to the Z ramigera biosynthetic thiolase The numbering of conserved active site water molecules... method, and the thermodynamic parameters were determined The reproducible binding constants were derived from at least two independent measurements Acknowledgements The authors wish to acknowledge the excellent support of the protein crystallography beamlines of the EMBL-Hamburg Outstation ⁄ DESY The skillful technical assistance of Ville Ratas and discussions with Dr A M Lambeir (University of Antwerp,... supplementary material is available: Fig S1 Electron densities of ligands in the crystal structures This supplementary material can be found in the online version of this article Please note: Wiley-Blackwell is not responsible for the content or functionality of any supplementary materials supplied by the authors Any queries (other than missing material) should be directed to the corresponding author for the. .. (1967) The hen egg-white lysozyme molecule Proc Natl Acad Sci USA 57, 483–495 36 Henderson R (1970) Structure of crystalline alphachymotrypsin IV The structure of indoleacryloylalpha-chyotrypsin and its relevance to the hydrolytic mechanism of the enzyme J Mol Biol 54, 341–354 37 Mattos C, Rasmussen B, Ding X, Petsko GA & Ringe D (1994) Analogous inhibitors of elastase do not always bind analogously Nat . The sulfur atoms of the substrate CoA and the catalytic cysteine are required for a productive mode of substrate binding in bacterial biosynthetic thiolase, a. for AcAc -CoA. In the crystal structure, the binding modes of the reactive sulfur moieties of SPP and CoA are highly similar, explaining the reactive nature